Resumo

VELAZQUEZ-ZAPATA, Juan Alberto  e  DAVILA-ORTIZ, Rodrigo. Uncertainty Related to Processed Gridded Meteorological Data: Implications for Hydrological Modelling. Ing. invest. y tecnol. [online]. 2017, vol.18, n.2, pp.199-208. ISSN 2594-0732.

Spatial interpolation is a procedure for estimating the value of a variable of interest at unsampled sites within an area covered by existing observations. The output of spatial interpolation is an integrated data set in which meteorological data are arranged along an evenly spaced matrix (gridded data). This work evaluates the uncertainty related to meteorological gridded data in the simulation of daily streamflow over two Mexican basins. The use of gridded data is an alternative to direct observations in those Mexican regions with low density of gauging stations. First, two meteorological data sets (observed and processed gridded data) were compared. Results show that gridded data underestimate precipitation, maximum and minimum temperature, despite the relative good agreement in the annual cycle for the latter variable. Second, the lumped conceptual rainfall-runoff model GR4J was fed with meteorological data from both data sets in order to evaluate the error that gridded data translate to simulated daily streamflow. Results show that the hydrological model can be calibrated with both data sets, leading to a good performance for medium and high flows in terms of the Nash- Sutcliffe efficiency coefficient; nevertheless, low flows are overestimated when gridded meteorological data are used. The analysis of the GR4J optimized parameters shows that the hydrological model increases the contribution of groundwater exchange to compensate for the underestimated precipitation, leading to a misrepresentation of the hydrological response of the study basins. All in all, gridded processed meteorological data should be evaluated before its use on hydrological risk assessment and climate change impact studies on water resources.

Palavras-chave : uncertainty; precipitation; hydrological modelling; GR4J.